It is considered that most cognitive processes emerge not solely in the brains but in their dynamic interactions with the environment associated with bodily behaviors. Based on this idea of the embodied cognition we attempt to describe the essential mechanism of cognitions coupled with sensory-motor systems using dynamical systems language. For this purpose, we are conducting interdisciplinary studies by taking the synthetic approach with neural network modeling and robotics experiments and the empirical approach including psychological experiments, electro-physiology and brain imaging. We focus on the following problems.
First, we focus on the problem of how behavior schemata or primitives can be organized in order to generate or recognize various and complex behavior patterns. More specific questions are that (1) whether they are represented locally or distributedly in networks, (2) how they can be represented in a hierarchical way. We propose a novel model and analyze its characteristics by conducting robotics experiments.
Second, we work on the problems of symbols and language. Although in the conventional cognitive science symbol representations and their manipulations are assumed as apriori, a crucial question is that how such symbol systems can be realized in the neuronal dynamical systems in brains. It is also asked how they can be acquired as grounded to the sensory-motor experiences. We study these problems by taking a synthetic approach using neural net modeling and robotics experiments.
Third, we conduct electro-physiological experiments of rats during their reward-based behavior and Bengalese Finches during their song learning for the purpose of exploring the essential neuronal mechanisms utilizing their hierarchical structures. We also conduct brain imaging using fMRI associated with psychological behavior analysis focusing on the human sequence learning mechanisms. The finding obtained in these empirical studies would provide novel ideas to the modeling studies mentioned above.
Forth, we study the development of interactions between robot and human in the processes of their co-adaptations. It is interesting to conduct numerical analysis in the robot side and phenomenological analysis in the human side simultaneously. It is expected that such studies will contribute to the developments of entertainment robots as well as to the creations of novel interfaces between human and machines including interactive arts.
Finally, we conduct basic studies for complex adaptive systems in order to gain theoretical understanding of adaptive behaviors in an open environment.